Alain-Michel Chomnoue Nghemning

To avoid performance issues and reduce costs, the API rate limiter allows a company to avoid too many calls to a given API.

The limits are configurable, and the rate limiter should throttle AIP calls in the following situations:
1. Too many requests within the same time window from a client
2. Too many requests from a specific client on a per month basis
3. Too many requests across the entire system

The application is built to run in a distributed environment and has been designed to scale effectively.

A serverless API that collects tweeter posts tagging targeted companies, runs a sentimental analysis on them, and computes their NPS (see Medallia.com/net-promoter-score/#). It runs on AWS, involving cloud services like Lambda, Firehose, S3, DynamoDB, Comprehend, and the NPS score for telcos (Social-nps.herokuapp.com/dashboard). On top of that, I built a dashboard with Python and the Bokeh library and deployed it on Heroku to display the computed NPS scores.

This visualization shows the evolution of Ebola cases in West Africa from August 2014 to May 2015, as reported by the WHO.

The user can select an indicator and a date to see the status on the map. He can also choose to replay the animation for the chosen indicator.

I chose to show the affected country on a map with the orange color differentiating them. Only the affected countries are named to let the user focus on them.

A red circle is used to show the value of an indicator for each country. I chose a logarithmic scale for the circle radiuses so that with just a few cases, countries' values can be seen on the map.